You are on page 1of 9

Introduction

Hydropower is a renewable energy resource that can produce electricity without


using fossil fuels. The environmental consequences of hydropower are related to the
disturbances of natural ecosystems due to damming, lowering of the water level,
altered water flow and new infrastructure. Norway is in an unique position due to
the fact that a large part of the domestic energy consumption and power demanding
industries, are fueled by electricity. Maintaining and developing the hydropower
industry is without a doubt crucial, especially in this day and age where climate
changes strongly encourage the use of green technologies.
The Pelton turbine absorbs energy from a high-speed water jet and is characterized
by high heads and small volume flows. Even though this is a well established
turbine technology, there are many unanswered questions regarding design and
optimization. Thus, further development is still relevant today.
In the autumn of 2010 MSc student, Kristine Gjster, began the process of developing a Francis turbine design software called Khoj. [10] Peter Joachim Gogstad
continued the development of Khoj with his MSc thesis. [11] Hopes are that more
people will be able to conduct experiments, and participate in the development of
the Francis turbine.
Khoj is a MATLAB application with a user friendly GUI that produces a geometry
output readable by Pro/ENGINEER and ANSYS CFX (A. CFX) for mechanical
design and hydraulic analysis, respectively. A similar software has been requested
for the Pelton turbine, but the lack of an open Pelton bucket design has prevented
this development.
Doctoral candidate, Bjrn Winther Solemslie, is currently in the process of designing a new reference Pelton turbine on behalf of NTNU. A new design provides the
opportunity to develop a Pelton design software, such as Khoj. This thesis will
initiate the development of a Pelton CFD model, that will be a part of a Pelton
design software in the future.
Numerous CFD analyses on the flow in Pelton buckets have previously been carried
out, but apparently a transient analysis of a free surface water jet interacting with
rotating Pelton buckets, has not yet been attempted at NTNU. The main objective
of this thesis will be to perform the latter, and validate the torque measurement
1

produced by the CFD model. To ensure that this work will benefit future students
at the Waterpower laboratory, a detailed procedure for CAD modeling, meshing
and A. CFX setup will also be created.
DynaVec, a company based in Trondheim, has experience with CFD analyses on
Pelton turbines. They offered their assistance and could provide a CAD model for
a bucket used in a model turbine. This turbine will be tested experimentally in the
Waterpower laboratory at NTNU, which will be a crucial part of the validation of
the
CFD
model.

Chapter 1

Previous Work
Previous research on Pelton turbines consist mainly of analytical, numerical and
experimental studies. This section will briefly highlight the research of interest to
this thesis.
Early on, professor H. Brekke analytically developed a method to reconstruct the
path of water particles on their relative momentum equations in the bucket. [3]
Brekke [12] based his work on a graphical particle trace method developed by
Henrik Christie in 1918. Hana [13] investigated and partially improved this method.
The numerical analyses of the free surface flow in a Pelton bucket have generally utilized commercial CFD tools with mesh-based solvers. Another approach is
the mesh-free Lagrangian method which completely eliminates the difficulties with
traditional mesh-based techniques. Smoothed-particle hydrodynamics (SPH) is a

The VIP model is an


to the
and commonly used VOF
ume of Fluid) model. The main difference between the
is
the VIP
calculates a necessary grid
between
steps resulting in a
of the
grid for every
step.
shows
the modeling issues in the VOF
is eliminated with the
model. Jet surface ripples and
up is discussed using
With Berntsens results, all flow features are
well up to about 2
away from the nozzle. The results are still accurate in some areas
down stream, but Berntsen concludes
the model needs further
in order to properly
for the inevitable jet break-up at large distances
the
Berntsens VIP model is something

could be useful and worth

promising Lagrangian method, but it is still in an early developing stage with few
published studies, which need to be correctly validated [2].
The experimental studies generally involve steady and transient pressure measurements, flow visualization and water film thickness measurements.

1.1

M. Hana

Morten Hanas PhD thesis "Numerical analysis of non-stationary free surface flow
in a Pelton bucket" [13], published in 1999, was the first study of its kind [3].
With the help of three commercial CFD-codes (RIPPLE, Flow-3D and CFX-4),
Hana ran both simplified 2D-simulations and transient 3D-simulations of the free
surface flow in a Pelton bucket. Three problem areas were highlighted: the free
surface, complex geometry and the relative motion of the jet and the buckets.
The performance of the different codes and methods were evaluated, and Hana

The VIP model is an


to the
and commonly used VOF
ume of Fluid) model. The main difference between the
is
the VIP
calculates a necessary grid
between
steps resulting in a
of the
grid for every
step.
shows
the modeling issues in the VOF
is eliminated with the
model. Jet surface ripples and
up is discussed using
With Berntsens results, all flow features are
well up to about 2
away from the nozzle. The results are still accurate in some areas
down stream, but Berntsen concludes
the model needs further
in order to properly
for the inevitable jet break-up at large distances
the
Berntsens VIP model is something

could be useful and worth

1.

G.

concluded that numerical calculations showed promising results and can replace
the graphical method when numerical calculations has been verified.

1.2

G. Berntsen

In 2003, Gotfred Severin Berntsen published his PhD thesis "Numerical analysis of
the two-phase Pelton jet flow using a single-phase model and analytical discussions
of the Pelton jet surface break-up phenomenon" [1]. With this thesis Berntsen
develops a single-phase CFD model called virtual interphase method (VIP) that
can be used to calculate any free surface jet where a high density ratio is present.

The VIP model is an


to the
and commonly used VOF
ume of Fluid) model. The main difference between the
is
the VIP
calculates a necessary grid
between
steps resulting in a
of the
grid for every
step.
shows
the modeling issues in the VOF
is eliminated with the
model. Jet surface ripples and
up is discussed using
With Berntsens results, all flow features are
well up to about 2
away from the nozzle. The results are still accurate in some areas
down stream, but Berntsen concludes
the model needs further
in order to properly
for the inevitable jet break-up at large distances
the
Berntsens VIP model is something

could be useful and worth

1.

G.

Figure 1.1: Normalized total pressure versus normalized radial distance. Dots
denote measured values and triangles denote numerical values. The similarities are
characterized
as
good.
[1]

The VIP model is an


to the
and commonly used VOF
ume of Fluid) model. The main difference between the
is
the VIP
calculates a necessary grid
between
steps resulting in a
of the
grid for every
step.
shows
the modeling issues in the VOF
is eliminated with the
model. Jet surface ripples and
up is discussed using
With Berntsens results, all flow features are
well up to about 2
away from the nozzle. The results are still accurate in some areas
down stream, but Berntsen concludes
the model needs further
in order to properly
for the inevitable jet break-up at large distances
the
Berntsens VIP model is something

could be useful and worth

1.

1.3

L.

A. Perrig

In 2007, Alexandre Perrig publish his comprehensive PhD thesis "Hydrodynamics


of the Free Surface Flow in Pelton Turbine Buckets" [2]. Perrig studied the flow in
the buckets with 4 experimental and numerical approaches: (i) Unsteady onboard
wall pressure measurements. (ii) High-speed flow visualizations. (iii) Onboard
water film thickness measurements. (iv) CFD simulations. He compared the 2Phase Homogeneous Model and the 2-Fluid Model, and concluded that the latter
one was the most accurate. However, the accuracy of the 2-Phase Homogeneous
Model decreases significantly with the successive time steps.

Figure 1.2: Details of Perrigs computational grid. [2]


The 2-Fluid Model is recommended and should be run with a large number of
iterations to reach complete convergence. Perrig states that the main drawback
associated with the classic CFD approaches is the need for refined meshes in the
whole domain, while the region of interest for any give time step only represent
a fraction of the domain. As this dramatically increases the computational cost,
accurate numerical simulations of a full-scale facility appears highly difficult.
Perrigs work is relevant to this thesis because his numerical CFD simulation setup
is very similar to what the author is trying to achieve. Experimentation with the
2-Phase Homogeneous Model and the 2-Fluid Model is also of interest.

1.4

L. Klemetsen

To the authors knowledge, the most recent norwegian study of the free surface flow
in Pelton turbine buckets is Lars Erik Klemetsens MSc thesis "An experimental
and numerical study of the free surface Pelton bucket flow" [3] published in 2010.
Klemetsen investigated a simplification of the free surface flow through a static
Pelton turbine bucket. Through experiments the pressure distribution throughout
the majority of the flow domain was obtained, as well as the location of the air-water
interface. Afterwards, these results were compared with numerical simulations

1.

L.

using two commercial CFD-solvers (Fluent 6.4 and CFX). This comparison showed
that the flow problem was practically independent of turbulence modeling and
mesh (as long as the grid resolution was kept above a certain lower limit). Gravity,
viscosity and surface tension could also be neglected due to the dominance of inertia
forces.
The use of mesh adaption algorithms at the water-air interface was an effective
improvement and the adjustments made to the velocity profile affected the results
the most. Klemetsen suggests the use of a dynamic mesh refinement for transient
full cycle simulations of rotating Pelton buckets due to the fact that the water
interface would change continuously. Also, a combined calculation of the nozzle
and the bucket profile would be a useful extension of his work.
Figure 1.3: Velocity boundary
conditions at the inlet. [3]

Figure 1.4: Definition of the velocity inlet in A.


CFX. [3]

All tough Klemetsens work is on a static bucket, it is of great value in order


to model the water sheet thickness, inlet boundary conditions and general CFD
simulation
setup.

1.

1.5

DynaVec AS

DynaVec is a company based in Trondheim, which specializes in designing, manufacturing, and installing pumps and turbines for environments with sediment erosion and corrosion problems. Their team consists of highly qualified personnel with
many years of experience in hydropower and renewable energy both nationally and
internationally. [14] For this thesis, their knowledge and experience of conducting
CFD analyses, were especially valuable.

DynaVecs Contribution
There was no Pelton CAD geometry with experimental data available at NTNU
when this thesis was initiated. Because experiments are very important in order
to validate a CFD model, it was a requirement to obtain experimental data of a
Pelton turbine in combination with a bucket geometry. In the autumn of 2011,
DynaVec agreed to share an early stage, noncommercial bucket geometry. This
geometry was used in the development of the method of modeling the fluid domain
and meshing this domain. Simulations, based on a method proposed by DynaVec,
were also conducted in order to see if the CFD model behaved within reason.
In March 2012 DynaVec provided the author with a new bucket geometry of a
model turbine, which was tested in the laboratory in April 2012. The CFD analysis
presented in this thesis is based on this geometry, and because it is commercial, all
results are normalized.